Objective: To investigate the internal construct validity of the Rivermead Post-Concussion Symptoms Questionnaire (RPQ) by Rasch analysis of data from a national cohort of patients with mild traumatic brain injury.
Methods: Data collected at 3 months after mild traumatic brain injury from 2,523 patients were analysed using the partial credit model, describing rating scale structure, local dependency, age and gender differential functioning, dimensionality and model fit.
Results: Categories did not work in a consistent manner; however, collapsing of Categories 1 and 2 yielded ordered thresholds. Local dependency of items was present and 2 item pairs were combined. There was no differential item functioning by gender or age. The Rasch factor explained 47.7% of the variance and the first contrast explained 12.4% of the unexplained variance (eigenvalue 1.9). Further analysis indicated 3 or more dimensions. Person measure had a mean of -2.16, showing poor targeting of persons to items. Person reliability was 0.71 and person separation index was 1.56.
Conclusion: According to this Rasch analysis of data from a representative sample of mild traumatic brain injury, the RPQ may not be optimal for this population. Even after reducing the number of categories and collapsing items with local dependency, unidimensionality was not reached, which argues against summation of a total score. However, the scale is unbiased for gender and age. :